Speaker verification is the process of verifying the identity of a speaker based upon an analysis of a sample of his or her speech using previously saved information. In general, speaker verification or authentication consists of making a determination as to whether the identity of a speaker is, in fact, the same as that being claimed. Some applications of speaker verification include, for example, access control for a variety of purposes, such as for telephones, computer networks, databases, bank accounts, credit-card funds, automatic teller machines, building or office entry, etc. Providing automatic verification of a person's identity based upon his or her voice can be implemented more conveniently and in a less costly manner than many other known biometric methods such as, fingerprint analysis or retinal/iris scanning. Further, unlike other biometric solutions, speaker verification is non-intrusive and may be implemented from a remote location. In addition to added convenience, speaker verification of user identity substantially increases security of phone-based communication systems over conventional authentication systems such as those using numeric PIN codes or passwords.
In general, speaker verification is performed based upon previously saved information which, at least in part, represents particular vocal characteristics of the speaker whose identity is to be verified. Specifically, the speech signal which results from a speaker's “test” utterance (i.e., an utterance offered for the purpose of verifying the speaker's identity) is analyzed to extract certain acoustic elements of the speech signal commonly referred to as a voice print. The previously uttered speech which is used for comparison purposes most commonly, but not necessarily, consists of a number of repetitions of the same word or phrase as the one which is to be spoken as the “test” utterance. In any case, the previously uttered speech is referred to as “training” speech, and it is provided to the system as part of an “enrollment” session.
In use, identified acoustic elements extracted a user's input voice signal are compared against the voice print identified from the test utterance. A determination is then made indicating whether the speaker is who they claim to be.
Unfortunately, speaker verification systems proposed to date require significant capital and infrastructure expense on the part of the entity desiring system implementation. In addition to procuring or developing software systems for establishing and comparing voice print data from enrolled participants, the voice prints generated during enrollment of users must be maintained in manner facilitating efficient utilization. Such limitations render speaker verification systems unusable to a wide-variety of otherwise interested entities.
Accordingly, there is a need in the art of speaker verification systems for a speaker verification system which reduces expense and infrastructure requirements for participating entities.